Poet Technologies Live Demo’s Three Products at OFC 2026

Poet Technologies is grabbing attention in the Photonics hardware field vital to AI Data Center Scaling.

There's a difference between a photonics story… and a photonics company. At OFC 2026, POET $POET Technologies tried to prove it's becoming the latter. The consensus is they succeeded.

For years, POET has been grouped into the same bucket as every other optical startup: big vision, elegant architecture, and timelines that always seemed just out of reach. This week felt different. Not because of what they said—but because of what they showed.

Three products. Live demos. Real hardware.

Blazar: The Cost Curve Bet Poet Launches Blazar External-Cavity Laser

The headline grabber was Blazar, POET's external light source (ELS) platform. The pitch is simple: break the traditional laser model.

Instead of embedding wavelength selection inside expensive Indium Phosphide (InP) chips, POET moves that function onto its Optical Interposer. The implication is massive—smaller InP chips, more dies per wafer, and potentially a structural collapse in laser costs.

That's the bet

POET is effectively saying the industry has been building lasers the wrong way—and that its architecture proves it can deliver similar performance at materially lower cost.

If that's true, it doesn't just compete. It resets pricing power across the optical stack.

But here's the reality: this is still a scaling story, not a finished one. Cost advantages don't exist until they show up in volume production. Until then, Blazar is a thesis the market is watching—not yet underwriting.

Starlight: Where It Starts to Get Real

Poet Demos Starlight Gen 2 at OFC Conference LA

If Blazar is the vision, Starlight is the bridge to commercialization.

This is POET's next-generation ELS platform targeting the emerging ELSFP standard, and more importantly—it's built on top of the company's Lightbar architecture, the closest thing POET has to a production-ready foundation.

At OFC, Starlight wasn't a concept. It was running inside an optical engine, demonstrating multi-channel, multi-wavelength capability in a format that actually maps to how customers deploy hardware.

That's a shift. Not a lab demo. Not a render. A system.

Still early—but no longer theoretical.

Teralight: The Quiet Power Move

Then there's Teralight, the 1.6T optical engine co-developed with Mitsubishi Electric. This is where things get interesting.

Most 1.6T designs rely on up to eight lasers. Teralight cuts that roughly in half using a 2×200G EML approach. In a world where InP supply is tight and costs are rising, that's not a small optimization—it's a strategic advantage.

Fewer lasers. Lower cost. Less complexity.

It also picked up a Lightwave Innovation Reviews score of 4.5/5, which in this industry actually means something.

The Part Nobody Should Ignore

Here's the part the market tends to get wrong with companies like POET:

Demos are not revenue.

The photonics graveyard is full of companies that worked perfectly… right up until they had to manufacture at scale, meet reliability standards, and survive customer qualification cycles. POET hasn't cleared that hurdle yet.

But what changed at OFC 2026 is this:

We're no longer debating whether the technology works.

We're debating whether it can scale.

Why This Matters Now

This isn't happening in a vacuum.

AI infrastructure demand is exploding. Optical interconnect is becoming a bottleneck. Power, cost, and density are now the constraints—not compute.

If POET's architecture delivers even part of what it promises:

— Lower cost per bit

— Reduced power consumption

— Less reliance on constrained materials

…it doesn't just win deals. It fits exactly where the industry is heading.

Bottom Line

OFC 2026 didn't prove POET is a an unequivocal winner, but it eliminated the biggest reason to ignore it.

The technology is real. The products exist. The demos worked.

Now comes the only question that matters:

Can they actually build it at scale—and get paid for it?